MEMS resonators are now being developed for use in frequency-specific applications, such as oscillator references and highly-selective bandpass filters. These applications require that the resonator possess a specific resonance frequency. For example, in the case of an oscillator that serves as part of a clock circuit, it is important that the resonator vibrates at a specific frequency. In the case of a filter, a resonator must likewise vibrate at a particular, targeted frequency to generated passband to selectively pass or reject a signal as a function of frequency.
Due to the vagaries of manufacturing, the measured resonance frequency of a resonator is typically different from its targeted value. Variations of about 5 percent are typical. As a consequence, a newly-manufactured resonator often needs to be tuned to adjust its resonance frequency. Tuning of a resonator is akin to tuning a piano, although the techniques used are quite different. The process of tuning a resonator is called “trimming.”
Frequency trimming is well-known. Indeed, it is commonplace to trim the resonance frequency of piezo-electric components (e.g., crystals, etc.), and resonators, oscillators, and clocks that incorporate them. Unfortunately, the techniques that are used for frequency trimming these devices are not well adapted for use with MEMS resonators.
For example, laser trimming has traditionally been used to trim the resonance frequency of crystals. But laser trimming has not been considered to be a viable technique for use with MEMS resonators since they are typically much smaller in size than their crystal counterparts. Also, it is desirable to trim a MEMS resonator after it has been packaged, but doing has not been considered to be feasible via laser. (See, e.g., Joachim et al., “Characterization of Selective Polysilicon Deposition for MEMS Resonator Tuning,” J. MEMS, v(12), no. 2, pp. 193–200 (April 2003) at p. 193; Lin et al., “Microelectromechanical Filters for Signal Processing,” J. MEMS, v(7), no. 3, pp. 286–294 (September 1998) at p. 293; U.S. Pat. Nos. 6,600,389 at col. 1, lines 33–39, and U.S. Pat. No. 6,570,468 at col. 1, lines 27–34).
Other techniques that have been used to trim piezo-electric components include removing mass by polishing or adding mass by depositing a resin (see, e.g., U.S. Pat. No. 6,604,266 at col. 1, lines 20–49). These techniques are not well suited for trimming MEMS resonators either. In particular, the very small size (micron and even submicron size) of MEMS resonators makes polishing and selective deposition difficult. Furthermore, the spring constant of the resonator is very sensitive to beam thickness. Variations in the spring constant, such as can be caused by removal or addition of material, will affect the quality factor, Q, of the resonator.
Since frequency-trimming techniques that have been used for piezo-electric components are not readily adapted for use with a MEMS resonator, new frequency-trimming techniques have been developed.
In a first frequency-trimming technique that is useful with MEMS resonators, the resonance frequency of a resonator is changed by modifying a structure that supports the resonator. See, U.S. Pat. No. 6,570,468. In this technique, the resonance frequency of a resonator is altered by changing the stiffness of a supporting structure. According to the patent, the stiffness of the supporting structure is modified by forming notches therein or by adding material thereto.
In a second approach to the problem of tuning a MEMS resonator, a structure having a plurality of beams of variable length is formed. See, U.S. Pat. No. 6,600,389. According to this approach, the variation in beam length results in a difference in resonance frequency between the shortest and longest beam that is sufficient to account for the typical variation (due to manufacturing tolerances) in the resonance frequency of a resonator. This patent also discloses that the increment in resonance frequency between adjacent beams is smaller than the targeted frequency variation tolerance of the desired resonator. Therefore, one of the beams will be qualified to serve as the desired resonator. Once that beam is identified, the other beams are disabled.
The techniques described in U.S. Pat. Nos. 6,570,468 and 6,600,389 are performed before the resonator is packaged. But typically, resonators must be operated under high vacuum conditions. To the extent that the techniques described in U.S. Pat. Nos. 6,570,468 and 6,600,389 are not performed under high-vacuum—the environment of a packaged resonator—there will be uncertainty as to the amount of frequency trimming that is required.
Furthermore, it is known that the resonator packaging itself can affect the resonance frequency of a resonator (see, e.g., Lin et al. at p. 293). In other words, the resonance frequency of a resonator can differ before and after encapsulation in a package. Since the first and second techniques discussed above and described in U.S. Pat. Nos. 6,570,468 and 6,600,389 are conducted before the resonator is packaged, there is, again, uncertainty as to how much frequency trimming is required.
In a third approach, which is referred to by its inventors as “localized annealing” or “filament annealing,” the resonator is trimmed after it is packaged. According to this approach, voltage pulses are applied to the resonator. This causes filament-like heating of portions of the resonator, which results in frequency trimming and improvements in Q. See, Wang et al., “Frequency Trimming and Q-Factor Enhancement of Micromechanical Resonators Via Localized Filament Annealing,” Dig. Tech. Papers, v(1), 1997 Int'l Conf. Solid-State Sensors and Actuators, Chicago, Ill., pp. 109–112 (Jun. 16–19, 1997).
While the third technique avoids the drawback of the first two approaches (i.e., trimming before packaging), it has some other deficiencies. In particular, the degree of trim control and Q control is very dependent upon the method used to dope the resonators in addition to other process-related variations.
As a consequence, there is a need for a method for trimming the resonance frequency of MEMS resonators that avoids at least some of the problems of the prior art.